Anomaly Detection tells you when your online conversations start behaving unusually — before you would notice it on a dashboard. Mentionlytics watches each tracker's mention activity against its own recent baseline, and emails you when something departs from it.
Anomaly alerts live with your other email notifications, in Settings → Notifications, under the heading Anomaly notifications in the Email Notifications panel.
What counts as an anomaly
In the context of social and online data, anomaly detection can flag:
Sudden spikes or drops in the volume of mentions or engagement
Unexpected sentiment changes (e.g., a rapid swing from positive to negative mentions)
Unusual traffic sources (e.g., mentions suddenly coming from an unfamiliar location or demographic)
What an alert contains
Anomaly alerts are produced by SIA, the Social Intelligence Advisor, which analyses your mention data continuously rather than on a schedule. Each alert gives you enough to decide whether to act without opening the app:
Mention volume chart
Shows when volume began to leave its normal range, so you can see the size and shape of the change against the expected baseline.
AI summary
Names the topic or event most responsible for the unusual activity, so you understand the cause without reading through every mention.
Create an anomaly notification
Go to Settings → Notifications.
In the Email Notifications panel, click +.
Change Notification type from New Mentions to Anomaly Detection. The form changes to the anomaly settings.
Fill in the settings below and click Save.
The new notification appears under Anomaly notifications. Its Frequency, Attachment and Filter Preset columns stay empty, which is expected — an anomaly alert has no schedule.
Customize AI Anomaly Detection Alerts
Through the Settings -> Notifications tab, you can customize everything:
Send to
Which users receive the alert.
Mention Tracker/Keyword
The trackers or keywords to watch for unusual activity.
Selected channels
Which sources to include. All by default — narrow it to leave out channels that are noisy or irrelevant to what you are watching.
Minimum hours between alerts
How long to wait after one alert before another can be sent, so a single event doesn't produce a run of emails. The default is 3.
Alert Threshold
The minimum increase in mentions that counts as an anomaly, measured against the tracker's recent average. The default is 20. Raise it for high-volume trackers where 20 extra mentions is an ordinary afternoon.
Check for negative mentions only
Count only negative mentions toward the threshold. Turn this on when your priority is reputation rather than reach — you will hear about criticism spreading, not about a popular post.
Sensitivity
How readily a change is treated as an anomaly. Toward High, smaller deviations qualify and you get more alerts; toward Low, only pronounced ones do. Start in the middle.
AI Summary Language
The language of the AI-written summary in the alert, so a team reading in Greek or Spanish gets it in their own language.
Additional e-mails (optional): CC / BCC
Send the alert to addresses outside the workspace — an agency contact, or an on-call address. CC recipients are visible to everyone on the email; BCC recipients are not.
Tuning
Threshold and sensitivity trade early warning against noise, and they interact: a low threshold with high sensitivity will alert on ordinary daily variation.
Too many alerts you don't act on? Raise the threshold first. You will keep detecting real events, just a little later. Lower the sensitivity only if raising the threshold isn't enough.
Nothing at all, ever? Your threshold is probably above your tracker's normal range of variation. Halve it and wait a week.
Repeated alerts about one incident? That's the alert gap, not the threshold. Raise Minimum hours between alerts.
Give any change a week of real traffic before judging it. Baselines are built from recent activity, so a tracker created yesterday has little to compare against.